In “The Terminator” series of movies and graphic novels, Skynet’s AI had evolved and became self-aware. It realized that it can be self-sufficient and that humans are dispensable and no longer needed for its growth and survival. Luckily for us, IT equipment and server workloads do not have that level of artificial intelligence or feelings (at least not yet), and they still need us as much as we need them. Being left out of the mission-critical group will not influence their performance, nor will they harbor any resentments. So, we’re safe for now.
Prioritizing or grouping IT and server workloads into mission-critical and non-critical categories is paramount to running a lean, efficient data center. Prioritizing can preserve power, cooling, and space capacity. Furthermore, workload shedding can be orchestrated in an organized fashion in the event of a massive power loss. Currently, in the real world, Data Center Operators and IT Managers often coordinate very minimally when distributing equipment in data centers, especially in mixed-use environments. Traditionally, mission-critical types of equipment and workloads, such as production servers, are housed in racks or cages that have redundant power, cooling, and connectivity. In contrast, non-mission-critical workloads are spread out in various places where there is extra or just enough space for power and cooling systems. There is nothing wrong with this approach if power utilization and optimization are not important to the facility manager. Chances are, it is.
In fact, The Uptime Institute recently published an article that highlights the advantages of planning out mixed-use data center IT distribution between production and non-production workloads. This is a good first step to help reduce human errors. However, it only applies to IT that is just being built out. With existing infrastructure, this still requires a lot of human intervention to physically move equipment, consuming resources and real estate as well as time. Moreover, this process can be especially vulnerable to human error. Automating IT to preserve mission-critical operations is key to making this vision a reality.
Software-Defined Power (SDP) can help data center operators achieve this goal and more, without monopolizing resources and/or introducing downtime caused by human errors. SDP provides sophisticated levels of control and automation that can be defined by the data center operators. With prudent planning for power and cooling based on a well-specified set of policies, the facility manager can more effectively:
- Distribute power between servers, racks, rows, cages, etc. through existing smart hardware without the need to move physical equipment.
- Greatly reduce the chance of human error in operations
- Create additional space and power capacity by unlocking trapped power saved for redundancy
- Empower operators with automation tools and insights on power profiles
- Identify and designate mission-critical and non-critical workloads and applications for workload shedding through SDP policies
- All while maintaining resiliency and redundancy needed for mission-critical workloads
On top of that, operators can leverage a plethora of other mechanisms from VPS’ ICE platform that can suit their needs such as:
- Addressing short-term high capacity power bursts
- Leveling out occasional power demands through peak-shaving
- Transferring workloads to other racks, rows, or even other data centers, through Workload Orchestration
- Shedding additional power consumption at the server/CPU level via Node Capping
- Offering additional available power to other users/customers that can lock-in that option for a certain time frame
- Improving visibility of their power consumption to make intelligent power predictions based on historical data through Machine Learning
Combining proper planning for mission-critical and non-critical IT, using ICE can stave off unnecessary data center expansion that can cost from $1.8M to $38M depending on size.
With software-defined power, facility managers can say “Hasta la vista” to hindering mission-critical operations with non-production IT and confidently say “I’ll be back” (in control, that is).